The most time/cost optimal solution is probably to buy 32 or 64 gigs of ram. That'll still be slow but most people are already half way there.

There is a neat potential speedup here for the case where the bandwidth to your model weights is the limiting factor.

If you have a guess what the model will output, then you can verify that your guess is correct very cheaply, since you can do it in parallel.

That means there is the possibility to have a highly quantized small model in RAM, and then use the big model only from time to time. You might be able to get a 10x speedup this way if your small model agrees 90% of the time.

Why 3 words per minute as opposed to second? Is that a typo? If you have enough RAM (but not VRAM), does it basically become limited by the PCIE lanes? So for the 112GB model with a Gen 5 GPU (64 GB/s PCIE bandwidth) that would be roughly 2 seconds per word right?

I just tried this on the 7B model. Steady state single threaded CPU performance of 23 seconds/token on a Ryzen 5800x (I'm not sure why it's only using a single thread... usually these libraries automatically use more) and 14GB of ram. It used more than double that amount of ram while loading the model, and the first token took 183 seconds (potentially it's doing more work to parse the prompt that I'm not measuring properly).